An examination of the "Texas ratio" as a bank failure model.
Jesswein, Kurt R.
INTRODUCTION AND BACKGROUND
So long as there have been banking institutions, there have been
banking failures. Whether by fraud and deceit or more commonly by poor
decision-making and risk management strategies, the banking industry has
periodically experienced severe downturns and suffered through the
failure and/or suspension of multiple institutions within very short
periods of time.
Although numerous at times, bank suspensions and failures prior to
1920 tended to be small in comparison to the ever-increasing number of
banks (Board of Governors, 1943). However, this all changed with the
coming of the roaring 20s and subsequent Great Depression years which
saw the number of banks across the country cut in half from over 30,000
in 1921 to around 15,000 at the time of the creation of the Federal
Deposit Insurance Corporation (FDIC) in 1934; over 9,000 banks suspended
operations from 1930 to 1933 alone (Board of Governors, 1943).
The introduction of the FDIC saw a marked change in this pattern as
the U.S. Department of the Treasury now had a mechanism in place to
assist banks that were experiencing difficulties. Although bank
suspensions continued at a pace of some fifty per year from 1934 to
1940, the FDIC was also assisting twenty-five to thirty banks per year
with the acquisition of failing institutions. This lead to a long period
of stabilization which saw double-digit bank failures only three times
in the ensuing four decades with most of the failures resulting in
purchase-and-assumption (P&A) transactions in which the FDIC helped
healthier institutions acquire most if not all of the failing
bank's assets and liabilities.
This all changed in the 1980s as deregulation of the banking
markets and increased volatility in the financial markets combined to
cause a significant increase in the number of troubled financial
institutions. As seen in Table 1, of the approximately 3,600 bank
failures that have been overseen by the FDIC since its creation, more
than 2,900 occurred between 1980 and 1993. Most of these transactions
involved purchase-and-acquisition transactions but the FDIC also became
involved with assisted acquisitions (AA), transactions in which it
provided direct financial assistance to institutions agreeing to acquire
failing institutions. There were over 500 of such AA transactions in
this time period with almost half of them occurring in 1988 during the
height (or perhaps better depth) of the fury of bank failures.
Passage of the Financial Institutions Reform, Recovery and
Enforcement Act of 1989 and subsequent Federal Deposit Insurance
Corporate Improvement Act of 1991 marked the next changes in the
handling of bank failures by the FDIC. There was a noted shift away from
assisted acquisitions to various purchase-and-acquisition transactions
as well as to direct payouts (PO), in which the FDIC simply paid off the
insured deposits and allowed to institution to fail.
As the 1990s progressed and the new millennium dawned, the banking
markets stabilized with relatively few financial institutions
failing--in fact there were NO failures in either 2005 or 2006--but this
would change. Significant upheavals in the financial markets towards the
end of 2007 and into 2008 and beyond have once again introduced an
increases amount of bank failures. This has created a situation in which
many bank customers and other interested parties are becoming
increasingly concerned about the health of their own financial
institutions. Sixty-two institutions have failed from the beginning of
2008 into early 2009. With such failures appearing to come with
increasing frequency, it is not unusual to find regular headlines such
as "If it's Friday, there must be a bank failing somewhere
across the country" (Ellis, 2009). Thus, there has been a renewed
interest in looking for ways to discover which financial institutions
were on the verge of financial failure.
REVIEW OF BANK FAILURE MODELS
Given the importance placed on banking institutions in the
operations of smoothly running economies, there have been varied
attempts to develop models to assist in finding those financial
institutions more likely to suffer financing hardships or worse. As
early as the 1930s we find examinations of the causes of bank failures
given the chaotic situation and widespread failures among financial
institutions during the late 1920s and early 1930s (Spahr, 1932).
Such studies all but disappeared until new groundbreaking work
focusing on the financial difficulties of industrial firms began to
appear in the late 1960s (Beaver, 1966; Altman, 1968). These studies
began to look to financial and accounting ratios as indicators of
financial distress through either univariate (Beaver) or multivariate
(Altman) models. Meyer and Pifer (1970) and Sinkey (1975) subsequently
developed models that examined financial difficulties of banks using
accounting and financial ratios more commonly associated with banking
institutions. For example, Sinkey incorporated such ratios as cash plus
U.S. Treasury securities to total assets, total loans to total assets,
provision for loan losses to total operating expenses, total loans to
sum of total capital and reserves, total operating expenses to operating
income, loan revenue to total revenue, U.S. government securities'
revenue to total revenue, municipal securities revenue to total revenue,
interest paid on deposits to total revenue, and other expenses to total
revenue in his study.
Subsequent studies tended to focus on the development and testing
of computer-based early warning systems (EWS) that might be used to
prevent bank failure or reduce the costs of failure. Such studies tended
to expand the quantitative analysis of the models (Kolari, Glennon, Shin
& Caputo, 2002; Wheelock & Wilson, 2000) or incorporate
efficient-market variables to examine stock price and interest rate
effects on the financial condition of financial institutions (Curry,
Elmer & Fissel, 2007; Purnanandam, 2007).
The primary bank regulatory institutions have also expanded their
efforts into refining and improving EWS models in the face of a
constantly-changing financial landscape. Examples of current systems in
use include the Federal Reserve's System to Estimate Examination
Ratings and Economic Value Model and the FDIC's Statistical CAMELS
Off-site Rating system and Real Estate Stress Test. (Cole & Gunther,
1998; King, Nuxoll & Yeager, 2006). Although differing in scale,
scope and purpose, these models continued to focus on the use of
financial variables to predict problem banks. One can simply contrast
the variables used by Sinkey with those used in the FDIC's SCOR
model: total equity, loan-loss reserves, loans past due 30-89 days,
loans past due 90+ days, nonaccrual loans, other real estate,
charge-offs, provisions for loan losses, income before taxes, volatile
liabilities, liquid assets, and loans and long-term securities, each as
a percentage of total assets (Collier, et. al., 2005).
On the other hand, a remarkably distinct yet extremely simplistic
tool has recently caught the fancy of many analysts in their attempts to
make sense of the turmoil that exists in the latter part of the first
decade of the 21st century. This tool, generally referred to as the
Texas ratio, focuses solely on only a couple specific accounting
variables that concisely summarize many of the credit troubles being
experienced by banks. The Texas ratio was first developed by Gerard
Cassidy and others at RBC Capital Markets in their analysis of Texas
banks experiencing difficulties during the troublesome 1980s (Barr,
2008). The ratio is calculated by dividing the bank's
non-performing assets (non-performing loans plus other real estate
owned) by the sum of its tangible equity capital and loan loss reserves.
Cassidy noted that the Texas ratio was a good indicator of banks likely
to fail whenever the ratio reached 100%. It has gained quite a bit of
notoriety in both the public media and in various areas of the
blogosphere, in part due to its simplicity and in part due to its
apparent success rate.
For example, one website, bankimplode.com, has attained a great
deal of notoriety since it began publishing its watch list of troubled
banks. This listing, based on publicly available bank call report data,
highlights all banks with Texas ratios greater than forty percent. The
website actually ranks the institutions using a separate measure, the
effective Tier 1 leverage ratio, but uses the Texas ratio as the
limiting variable. This effective Tier 1 leverage ratio will be
discussed later in the summary and conclusions part of the paper.
The FDIC itself maintains a watch list of troubled institutions.
However, its listing is not publicly available so speculation on which
institutions are on the list has led many to look towards measures such
as the Texas ratio to derive their own lists.
Based on the bankimplode.com watch list published after the third
quarter of 2008 we find that twenty-five of the fifty banks with the
highest Texas ratios had failed within the subsequent six months. In
fact, thirty-four of the forty-six institutions failing since the end of
the third quarter of 2008 were found somewhere on the bankimplode.com
watch list. Of the twelve banks not found on the watch list, one failed
without ever having submitted a third quarter call report, four had
Texas ratios just short of the artificial forty percent cut-off for
inclusion on the list, three were savings associations that submitted
financial reports to the Office of Thrift Supervision instead of the
FDIC, and one bank failed despite having a Texas ratio of only twelve
percent. The remaining three institutions not yet accounted for appear
to have had Texas ratios above forty percent but for some reason were
not included in the watch list.
Thus, it appears that there may be something behind this simple
measure for quickly assessing those financial institutions in serious
danger of failing. We are therefore left with examining the apparent
usefulness of the ratio and assess this usefulness relative to other
more sophisticated measures.
DATA AND METHODOLOGY
All data for the study were gathered from quarterly FDIC call
reports available through the Federal Reserve Bank of Chicago's
website at www.chicagofed.org. Our analysis focused on banks with total
assets between $20 million and $5 billion as the entire population of
banks failing since 2001 fall into this range. Note however that some of
the more newsworthy failures over the past two years were savings
institutions (IndyMac, Washington Mutual) and as such, were not included
in the study because their data are not included the data files
available from the Fed Chicago. For those interested, data on such
savings institutions are available through the FFIEC (Federal Financial
Institutions Examination Council) website at cdr.ffiec.gov/public; data
on commercial banks are also available at this website. And financial
data of credit unions can be found at the website of the National Credit
Union Association at www.ncua.gov.
Our study focuses on the time period encompassing all of 2008 and
in to the first four months of 2009 as there were only a handful of
failures in the years before 2008. The rapid deterioration of the
soundness and stability of so many financial institutions beginning in
2008 called for an examination of the most current data available.
We examine the situation surrounding bank failures occurring during
this time period by comparing data of failed institutions to the much
larger set of institutions that did not fail. We review how well the
Texas ratio has worked in terms of isolating those institutions more
likely to fail. We then attempt to discern any significant differences
between failing institutions and those that have not (as yet) failed.
Finally, we look to see if an expansion or modification of the Texas
ratio might be necessary to improve upon the basic model in terms of
providing more specific early warnings of bank problems.
RESULTS
As described earlier, the published watch list of banks with Texas
ratios greater than forty percent correctly identified seventy-three
percent (thirty-two of forty-eight) of the failing banks. And none of
the non-identified institutions had a Texas ratio less than twelve
percent. Based on this anecdotal evidence, the Texas ratio appears to
provide some much important insights.
Further examination shows that for the four quarterly periods
leading to the third quarter of 2008, the average Texas ratio increased
for failed and nonfailed banks alike. The average ratio for the small
group of banks that have failed in the past six months was 45 percent,
79 percent, 108 percent, and 181 percent, respectively. For the larger
group of over 7,000 banks that did not fail, the ratios were 9 percent,
11 percent, 12 percent, and 15 percent, respectively.
This leads us to examine in greater detail what the Texas ratio may
be measuring and whether that measure could be improved upon. The size
of the Texas ratio is essentially driven by the proportion of
nonperforming assets in a bank's portfolio and the bank's
concerns over future problem loans. The numerator of the ratio is
comprised of items that specifically represent assets that have gone bad
(nonperforming and/or foreclosed upon loans) and the denominator is in
large part affected by current and historical problems associated with
such assets (past credit losses that directly reduce the value of the
bank's equity and current credit problems that affect bank
profitability and the bank's ability to increase equity), and of
potential credit problems affecting the loss reserve account.
Because all credits are not created equal, a review of bank loan
portfolios may shed some light on specific items affecting the increases
in the Texas ratios of failed and nonfailed banks alike. For example,
banks are required to report results for a variety of different types of
credit including real estate construction and development, farmland,
residential mortgages (first and junior liens), home equity lines of
credit (HELOCs), multifamily residential properties, commercial real
estate, loans to depository institutions, to foreign governments and
official institutions, and to municipalities, loans to finance
agricultural production, commercial and industrial (i.e., business)
loans, various types of consumer loans, and lease financing. Few banks
have significant amounts of activity in all of the various sectors and
most only concentrate on small subsets.
As documented in Table 2, we find that there is a marked difference
in the lending portfolios of banks that have failed and those that have
not. For example, failed banks have a significantly higher percentage of
assets invested in real estate financing. However, this does not carry
over to all types of real estate financing. Failed banks have much
higher concentrations in construction and development loans. On the
other hand they have much lower amounts of secured lending such as for
first and second mortgages as well as for farmland and direct consumer
lending. And quite surprisingly based on recent media coverage, there is
very little difference between failed and nonfailed institutions in
terms of their exposures to either home equity lines of credit or to
commercial real estate. Thus, it would appear that rather focusing
solely on a single measure that captures all of the credit risks to
which banks are exposed, greater insights might be gained by expanding
or at least supplementing the Texas ratio with an examination of the
specific portfolio composition of a bank's risk exposure.
Note: Folded-F tests provide evidence that the variances for the
two groups are different. Therefore, the Satterthwaite t-test is
indicated. It provides a t statistic that asymptotically approaches a t
distribution. Wilcoxon z-scores are provided due to the potential of
having non-normal distributions, particularly in the small sample of
failed banks, and confirm the parametric results.
Additional insights might also be gained by examining credit
problems within each asset sector, particularly if specific sectors are
deemed to be more volatile or more likely to cause difficulties. Such
details (e.g., past due and nonaccruing amounts by asset sector) are
available from the data sources mentioned earlier and subsequent studies
of these data could provide important insights in future assessments of
the phenomenon of failing banks.
Another potential benefit associated with measuring the Texas ratio
is its ability to timely measure the potential for bank failures.
Although the Texas ratio appears to be a good indicator of bank problems
in the short term, one could argue that for such problems to arise to
such an extent as to cause the ratio to become excessive there would
likely be early warning signs. This in large part is the rationale
behind the Early Warning Systems used by the FDIC and Federal Reserve
System described earlier.
In Table 3, we examine the historical results of the key drivers of
the Texas ratio (nonaccruing loans, other real estate owned, and
allowance for loan losses). We find a significant demarcation between
failing and nonfailing banks in these measures, as well as the Texas
ratio itself, beginning at least three quarters earlier. Thus, even as
an early warning device, the Texas ratio appears to have some validity.
SUMMARY AND CONCLUSIONS
The Texas ratio has become a much publicized measure associated
with those banking institutions that are most likely to fail. But is it
truly a useful indicator? We have shown that it does appear to have some
merit. The intuition behind the ratio itself is solid and it can be
calculated with only minimum effort with readily available data.
However, that does not necessarily mean that it is a panacea for
all who may be looking for such a measure. For example, there can be
marked differences between types of loans and an individual bank's
exposure to specific types of lending. The Texas ratio includes only
institutional totals (total amounts of loans, nonaccruals, etc.) and
does not specifically examine loan portfolios. Certain types of loans
tend to have higher likelihoods of going into nonaccrual or default
status so banks making a higher proportion of those types of loans will
have higher Texas ratios and hence will be more prone to failure.
However, the Texas ratio, as currently defined, does not take into
account these differences in loan portfolios.
Furthermore, categorizing a loan as being in nonaccrual or default
status says little about the value of any collateral associated with the
loan and hence the actual amount of the loss given such a default.
Defaults on some types of loans may result in higher levels of loss, but
only in cases in which borrowers actually default. And the loans
themselves might have been quite profitable prior to any default,
allowing the bank to build up reserves against potential defaults to
help mitigate the seriousness of the loss.
One could also consider the opposite situation in which specific
forms of lending are not particularly profitable but also not considered
particularly risky. If no reserves are built up due to a previous lack
of profitability, only a modicum of credits going into could cause
significant problems.
One potential solution to this problem would be the development of
a companion or expanded measure. In fact, as mentioned earlier, a major
promoter of the Texas ratio measure, the analysts publishing through the
implode.com website, have themselves developed such a measure. In fact,
they use their complementary measure, the effective Tier 1 leverage
ratio, as their primary tool in ranking institutions most in danger of
failing, and use the Texas ratio itself as only a limiting variable in
comprising their watch list. The effective Tier 1 leverage ratio
attempts to estimate the impact on the capital of the bank (and hence
likelihood of bank failure) of actual losses expected on different types
of loans.
Although currently applied on a very ad-hoc basis, a measure such
as the effective Tier 1 leverage ratio measure could be made stronger
with greater availability of publicly-available data on the amounts of
loss given default experienced by different loan classes. By weighting
individual components of a bank's lending portfolio by those types
of assets more likely to cause actual losses and hence endanger the
bank's financial health, it can provide a more direct measure
rather than the one size fits all measure of the Texas ratio itself.
In conclusion, the rapid acceptance of using the Texas ratio to
examine the potential failure of banks has become a very interesting
phenomenon. The ratio is based on data that is readily available for any
and all types of financial institutions, involves only simple
calculations, and provides very straightforward output. This simplicity
is a key distinction from more rigorous models, including those found
elsewhere on the internet such as those provided by thestreet.com
(Weiss, 2009).
Although there is always a potential downside to providing simple
people with simple tools to assess very complex situations such as bank
failures, the use of a simple tool like the Texas Ratio can provide
individuals with a starting point from which more in-depth analyses of
the financial situation of banks can begin. To offer an analogy from the
books of Douglas Adams, it may not be the answer to "life, the
universe, and everything" (the answer to which is "42"),
but it brings us closer to understanding the types of questions that
need to be raised by those truly concerned with the financial health of
financial institutions. Given the rapidly increasing level of bank
failures, one can only presume that there will be a greater amount of
interest placed in this area, both in academia and among the general
population.
REFERENCES
Altman, E. (1968). Financial ratios, discriminant analysis and the
prediction of corporate bankruptcy. Journal of Finance, 23(4), 589-609.
Barr, A. (2008). Bank failures to surge in coming years. Retrieved
from MarketWatch, January 3, 2009 from
http://www.marketwatch.com/news/story/
bank-failures-surge-credit-crunch/story.aspx?Guid={2FCA4A0C-227D-48FE-
B42C-8DDF75D838DA}.
Beaver, W. (1966). Financial ratios as predictors of failure.
Journal of Accounting Research 5, 71-111.
Board of Governors of the Federal Reserve System. (1943). Banking
and Monetary Statistics 1914-1941. Retrieved February 3, 2009 from
FRASER (Federal Reserve Archival System for Economic Research) at
http://fraser.stlouisfed.org/publications/bms/issue/61.
Cole, R.A. & J.W. Gunther. (1998). Predicting bank failures: a
comparison of on- and off-site monitoring systems. Journal of Financial
Services Research 13(2), 103-117.
Collier, C., S. Forbush, D. Nuxoll, & J. O'Keefe. (2005).
The SCOR system of off-site monitoring: its objectives, functioning, and
performance. FDIC Banking Review 15(3), 17-32.
Curry, T.J., P.J. Elmer & G.S. Fissel (2007). Equity market
data, bank failures and market efficiency. Journal of Economics and
Business 59(6), 536-559.
Ellis, D. (2009). Get ready for a wave of bank failures. Retrieved
February 21, 2009 from
http://money.cnn.com/2009/02/20/news/companies/bank_failures/
index.htm?postversion=200902201.
King, T.B., D.A. Nuxoll & T.J. Yeager (2006). Are the Causes of
Bank Distress Changing? Can Researchers Keep Up? Federal Reserve Bank of
St. Louis Review 88(1), January-February, 57-80.
Kolari, J., D. Glennon, H. Shin & M. Caputo (2002). Predicting
large US commercial bank failures. Journal of Economics and Business
54(4), 361-387.
Meyer, P.A. & H.W. Pifer. (1970). Prediction of bank failures.
Journal of Finance 25(4), 853-68.
Purnanandam, A. (2007). Interest rate derivatives at commercial
banks: An empirical investigation. Journal of Monetary Economics 54(6),
1769-1808.
Sinkey, J.F. (1975). A multivariate statistical analysis of the
characteristics of problem banks. Journal of Finance 30(1), 21-36.
Spahr, W.E. (1932). Bank failures in the United States. American
Economic Review 22 (March supplement), 208-246.
Weiss, M. (2009). Dangerous unintended consequences: How banking
bailouts, buyouts and nationalization can only prolong America's
second great depression and weaken any subsequent recovery. Retrieved
April 16, 2009 from http://www.moneyandmarkets.com/files/documents/banking-white-paper.pdf.
Wheelock, D.C. & P.W. Wilson. (2000). Why do banks disappear?
The determinants of U.S. bank failures and acquisitions. Review of
Economics and Statistics 82(1), 127-138.
Kurt R. Jesswein, Sam Houston State University
Table 1: Summary of FDIC-Assisted Bank Failures (1933 - 2009)
P&A PO AA Other Total
before 1980 251 307 4 0 562
1980 7 3 12 0 22
1981 5 3 31 1 40
1982 25 8 85 1 119
1983 35 7 49 8 99
1984 62 5 23 16 106
1985 87 26 41 26 180
1986 98 25 42 39 204
1987 133 15 45 69 262
1988 165 7 238 60 470
1989 319 71 3 141 534
1990 291 44 1 46 382
1991 241 9 3 18 271
1992 153 12 2 14 181
1993 42 8 0 0 50
1994-2007 65 5 0 3 73
since 2008 52 2 5 3 62
Total 2031 557 584 445 3617
P&A = purchase-and-assumptions, PO = payouts,
AA = assisted acquisitions
Table 2: Differences in Credit Patterns: Failed vs. Nonfailed Banks
Failed Nonfailed
Banks Banks Satterthwaite
Type of lending Means Means t-statistic
(Percentage of total loans) N=37 N=7075 (Means)
Real estate 0.8119 0.6995 4.16 **
Construction & Development 0.3694 0.1111 7.51 **
Farmland 0.0289 0.0615 -2.94 *
HELOC 0.0316 0.0286 0.33
First Home Mortgage 0.1100 0.2195 -5.78 **
Second Home Mortgage 0.0097 0.0178 -4.88 **
Multifamily 0.0328 0.0212 1.30
Commercial 0.2295 0.2397 -0.48
Business 0.1168 0.1460 -1.60
Consumer 0.0215 0.0666 -7.08 **
Wilcoxon
Type of lending Z-score
(Percentage of total loans) (Medians)
Real estate 4.03 **
Construction & Development 8.02 **
Farmland -3.97 **
HELOC 0.27
First Home Mortgage -4.83 **
Second Home Mortgage -1.95 *
Multifamily 1.61
Commercial -0.29
Business -2.66 *
Consumer -5.95 **
* denotes significance at 5% level
** denotes significance at 1% level
Table 3: Historical Components of Texas Ratio: Failed vs. Nonfailed
Banks
Failed Banks Nonfailed Banks
Loan Statistic Means Means
(Percent of total assets) N=37 N=7075
Nonaccruing 0.1202 0.0148
Nonaccruing (-1 qtr) 0.0941 0.0126
Nonaccruing (-2 qtr) 0.0734 0.0108
Nonaccruing (-3 qtr) 0.0405 0.0088
Other real estate owned 0.0404 0.0051
Other real estate owned (-1 qtr) 0.0310 0.0041
Other real estate owned (-2 qtr) 0.0191 0.0034
Other real estate owned (-3 qtr) 0.0116 0.0028
Allowance for loan losses 0.0339 0.0138
Allowance for loan losses (-1 qtr) 0.0285 0.0135
Allowance for loan losses (-2 qtr) 0.0235 0.0134
Allowance for loan losses (-3 qtr) 0.0171 0.0130
Texas ratio 1.7951 0.1499
Texas ratio (-1 qtr) 1.0803 0.1257
Texas ratio (-2 qtr) 0.7926 0.1055
Texas ratio (-3 qtr) 0.4547 0.0885
Satterthwaite Wilcoxon
Loan Statistic t-statistic Z-score
(Percent of total assets) (Means) (Medians)
Nonaccruing 8.29 9.31
Nonaccruing (-1 qtr) 7.01 8.42
Nonaccruing (-2 qtr) 6.73 8.33
Nonaccruing (-3 qtr) 6.00 7.73
Other real estate owned 4.37 6.81
Other real estate owned (-1 qtr) 4.42 6.77
Other real estate owned (-2 qtr) 3.99 5.67
Other real estate owned (-3 qtr) 3.55 5.29
Allowance for loan losses 6.35 8.4
Allowance for loan losses (-1 qtr) 6.06 7.68
Allowance for loan losses (-2 qtr) 4.38 6.58
Allowance for loan losses (-3 qtr) 2.94 4.06
Texas ratio 6.65 9.88
Texas ratio (-1 qtr) 7.89 9.28
Texas ratio (-2 qtr) 7.45 9.52
Texas ratio (-3 qtr) 6.63 8.75
All variables significant at the 1% level
